Research Article

Cycloserine enantiomers are reversible inhibitors ofhuman alanine:glyoxylate aminotransferase:implications for Primary Hyperoxaluria type 1Mirco Dindo1,2,*, Silvia Grottelli1,*, Giannamaria Annunziato3, Giorgio Giardina4, Marco Pieroni3,Gioena Pampalone1, Andrea Faccini5, Francesca Cutruzzolà4, Paola Laurino2, Gabriele Costantino3,5 andBarbara Cellini1

1Department of Experimental Medicine, University of Perugia, Perugia, Italy; 2Okinawa Institute of Science and Technology Graduate University, 1919-1 Tancha, Onna-son,Okinawa 904-0412, Japan; 3P4T Group, Department of Food and Drug, University of Parma, Parma, Italy; 4Department of Biochemical Sciences ‘A. Rossi Fanelli’, SapienzaUniversity of Rome, Rome, Italy; 5Centro Interdipartimentale Misure G. Casnati, Parco Area Delle Scienze 23, University of Parma, Parma, Italy

Correspondence: Barbara Cellini (barbara.cellini@unipg.it) or Mirco Dindo (mirco.dindo@unipg.it)

Peroxisomal alanine:glyoxylate aminotransferase (AGT) is responsible for glyoxylatedetoxification in human liver and utilizes pyridoxal 50-phosphate (PLP) as coenzyme. Thedeficit of AGT leads to Primary Hyperoxaluria Type I (PH1), a rare disease characterizedby calcium oxalate stones deposition in the urinary tract as a consequence of glyoxylateaccumulation. Most missense mutations cause AGT misfolding, as in the case of theG41R, which induces aggregation and proteolytic degradation. We have investigated theinteraction of wild-type AGT and the pathogenic G41R variant with D-cycloserine (DCS,commercialized as Seromycin), a natural product used as a second-line treatment of mul-tidrug-resistant tuberculosis, and its synthetic enantiomer L-cycloserine (LCS). In contrastwith evidences previously reported on other PLP-enzymes, both ligands are AGT revers-ible inhibitors showing inhibition constants in the micromolar range. While LCS under-goes half-transamination generating a ketimine intermediate and behaves as a classicalcompetitive inhibitor, DCS displays a time-dependent binding mainly generating an oximeintermediate. Using a mammalian cellular model, we found that DCS, but not LCS, isable to promote the correct folding of the G41R variant, as revealed by its increased spe-cific activity and expression as a soluble protein. This effect also translates into anincreased glyoxylate detoxification ability of cells expressing the variant upon treatmentwith DCS. Overall, our findings establish that DCS could play a role as pharmacologicalchaperone, thus suggesting a new line of intervention against PH1 based on a drug repo-sitioning approach. To a widest extent, this strategy could be applied to other disease-causing mutations leading to AGT misfolding.

IntroductionAlanine:glyoxylate aminotransferase (AGT, EC 2.6.1.44) catalyzes the conversion of L-alanine andglyoxylate into pyruvate and glycine in liver peroxisomes, using pyridoxal 50-phosphate (PLP) as coen-zyme [1]. The deficit of AGT, as a consequence of inherited mutations on the AGXT gene, leads toaccumulation of glyoxylate that is eventually oxidized to oxalate, a metabolic end-product that precipi-tates as calcium oxalate (CaOx) stones in the kidneys and, in the most severe forms, in the wholebody. This rare condition is named Primary Hyperoxaluria Type I (PH1, OMIM 259900) [2,3]. PH1patients are currently treated by liver transplantation, a procedure coming with serious side effects, orby the administration of pyridoxine, a precursor of PLP effective in a minority (30%) of patients [4–6].

*These authors contributedequally to this work.

Accepted Manuscript online:3 December 2019Version of Record published:20 December 2019

Received: 9 July 2019Revised: 12 November 2019Accepted: 3 December 2019

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society 3751

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

AGT is homodimeric and belongs to the Fold Type I family of PLP-dependent enzymes. The two subunitsinteract through an extensive core interface, as well as through an N-terminal extension wrapped over theneighboring monomer. The active site is made up of residues coming from both subunits and PLP is covalentlybound through a Schiff base linkage with Lys209 [7]. The AGT catalytic mechanism is typical ofPLP-dependent aminotransferases and comprises two half-reactions [1]. In the first one, the α-amino group ofthe substrate L-alanine displaces the ε-amino group of Lys209 producing the external aldimine. Then, Lys209acts as a general base for the 1,3-prototropic shift generating a ketimine intermediate, which hydrolyzes topyruvate and pyridoxamine 50-phosphate (PMP) [8]. In the second half-reaction, glyoxylate binds to AGT–PMP and, through the same steps of the first reaction but in a reverse order, is converted to glycine regenerat-ing AGT–PLP.PH1 is an extremely heterogenous disease from a genetic point of view (∼200 pathogenic mutations are

known) [9]. Nevertheless, biochemical and cell biology studies agree upon the view that most missense muta-tions cause folding defects in AGT that can result in (i) an increased aggregation propensity, either in thecytosol or inside peroxisomes, (ii) a reduced stability of the dimeric structure, (iii) an increased susceptibility toproteolytic degradation, and/or (iv) a mislocalization of the protein to mitochondria for mutations co-inheritedwith the polymorphic P11L mutation that generates a putative mitochondrial targeting sequence. For thesereasons, PH1 is referred as a misfolding disease [10,11].The use of small molecules that bind a target protein and enable its proper folding, namely pharmacological

chaperones (PCs), is a promising strategy to treat misfolding diseases. PCs have been successfully applied totreat lysosomal storage disorders, phenylketonuria, tyrosine hydroxylase deficiency, cystic fibrosis, nephrogenicdiabetes insipidus [12]. In the case of enzymatic deficits, they are usually competitive inhibitors of the targetenzyme. This approach has been also explored in PH1. Aminooxyacetic acid is a well-known AGT inhibitorbehaving as chaperone to rescue some of the most common mutations leading to PH1, but its low specificitymakes it unsuitable for in vivo applications [13]. To identify potential PCs specific for AGT, we looked forknown inhibitors of PLP-enzymes approved for clinical use. Among them, cycloserine (CS) is a cyclic analog ofalanine or serine able to bind many PLP-dependent enzymes including transaminases, racemases and decar-boxylases [14–19]. The two enantiomers of CS show different properties. D-cycloserine (DCS) is a broad-spectrum antibiotic isolated from Streptomyces sp. strains used as a second-line drug in combination therapiesto treat tuberculosis, whose main target is alanine racemase [15,18]. DCS is also used in neurological studiesfor its ability to behave as an agonist of NMDA receptors [20]. L-cycloserine (LCS) is the synthetic enantiomerof CS commonly used as a regulator of lipid metabolism in cellular studies [21]. A common feature of LCS andDCS is that they are considered irreversible inhibitors of PLP-enzymes, because they form stable covalentadducts with the cofactor. In some cases, like in branched-chain aminotransferase, alanine racemase, and serinepalmitoyl transferase [14,15,17], both LCS and DCS behave as inhibitors, although with different kineticproperties.Using a combination of UV-Vis spectroscopy, enzyme kinetics, X-ray crystallography, and mass spectrom-

etry, here we explored the interaction of the two CS enantiomers with wild-type AGT and the G41R pathogenicvariant showing a folding defect leading to an increased tendency to aggregation and degradation [22,23].Surprisingly, we found that both LCS and DSC behave as reversible competitive inhibitors, although the stereo-chemistry of the ligand influences inhibition kinetics and mechanism. Studies in a cellular model demonstratethat DCS, but not LCS, behaves as PC for the G41R variant. The significance of the results obtained for thegeneral mechanism of inhibition of PLP-enzymes by CS, as well as for the possible future use of the CS chem-ical scaffold for the development of more specific PCs is discussed.

Materials and methodsMaterialsPLP, L-alanine, sodium glyoxylate, rabbit muscle L-lactic dehydrogenase (LDH), β-Nicotinamide adeninedinucleotide reduced form (NADH), isopropyl-β-D-thiogalactoside (IPTG), LCS, DCS and imidazole were allpurchased from Sigma. Ham’s F12 Glutamax medium and Zeocin were purchased from Invitrogen. Geneticinwas purchased from Gibco. Protease Inhibitor Cocktails were purchased from Roche. Water and organic sol-vents were purchased from Thermo Fisher Scientific. All other chemicals unless otherwise noted, were fromSigma–Aldrich or Wako.

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society3752

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

AGT expression and purificationEscherichia coli BL21 cells were transformed with the constructs pAGThisMa and pAGThisG41R-Ma encodingthe most common polymorphic form of AGT (major allele here called wild-type AGT) and the G41R varianton the background of the major allele, respectively. Expression and purification of the two proteins were per-formed as already reported [24].

Spectroscopic measurementsAbsorbance spectra were registered using a Jasco V-750 spectrophotometer using 1 cm path length quartz cuv-ettes at a protein concentration of 0.1–20 mM in 0.1 M potassium phosphate buffer (KP), pH 7.4. Wild-typeAGT and the G41R variant concentrations were determined using the previously determined molar absorptioncoefficient of 95 400 M−1 cm−1 at 280 nm [24]. PLP content was measured after treatment of the holoenzymewith 0.1 M NaOH using the molar absorption coefficient of 6600 M−1 cm−1 of free PLP at 388 nm.Fluorescence data were obtained using 1 cm optical quartz cuvettes in a Jasco FP-8200 fluorimeter with excita-tion and emission bandwidths of 5 nm. All measurements were performed in KP 0.1 M pH 7.4.

Activity assaysTo determine the kinetic parameters of the L-alanine/glyoxylate pair, purified proteins (0.1–0.2 mM) were incu-bated in the presence of 100 mM PLP in KP 0.1 M pH 7.4 at 25°C at increasing substrate concentrations and afixed co-substrate concentration. The reaction was stopped by adding TCA 10% (v/v) and pyruvate productionwas measured using a spectrophotometric assay coupled with LDH [25]. Data were fitted to the Michaelis–Menten equation.The AGT enzymatic activity in cell lysates was determined by incubating 100 mg of lysate with 0.5 M

L-alanine and 10 mM glyoxylate in the presence of 100 mM PLP for 20 (wild-type AGT) or 50 (G41R variant)min in KP 0.1 M pH 7.4 at 25°C. Upon addition of TCA 10% (v/v), pyruvate production was measured usingthe spectrophotometric assay coupled with LDH.

Inhibition studiesThe IC50 for LCS and DCS was determined by incubating each species in the presence of inhibitor at increasingconcentrations (from 1 mM to 50 mM) and then measuring the residual transaminase activity. Each mixturecontained 0.1 mM purified recombinant enzyme, 100 mM PLP, LCS or DCS, 10 mM glyoxylate, in KP 0.1 MpH 7.4. The reaction was started by the addition of 50 mM L-alanine.The inhibition constant (KI) of wild-type AGT and the G41R variant for LCS was determined by calculating

the kinetic parameters for L-alanine at various LCS concentrations (from 50 to 400 mM) and by performing aglobal (shared-parameter) fitting of the data using a competitive inhibition model.The slow-binding inhibition kinetics of DCS was evaluated using the progression curve method [26]. A solu-

tion of 0.1 mM wild-type AGT or 0.2 mM G41R variant in the presence of 100 mM PLP was added to reactionmixtures containing DCS (from 1 mM to 10 mM), L-alanine (250 mM), and glyoxylate (10 mM), in KP 0.1 MpH 7.4 and incubated at 25°C. At various times, aliquots were collected, the reaction was stopped by the add-ition of TCA 10% (v/v), and pyruvate produced was measured using LDH as coupled enzyme. Data for eachprogression curve were fitted to the integrated rate equation for slow binding inhibitors

[P] ¼ vst þ (v0 þ vs)(1� exp(�kobs)= kobs) (1)

where, v0 represent the initial rate, vs the steady-state rate, kobs the apparent first-order rate constant character-izing the formation of the steady-state enzyme–inhibitor complex. The obtained kobs values were further ana-lyzed for a one-step association mechanism

kappobs ¼ kappon [I]þ koff (2)

where, koff and kon are the dissociation and association rate constants, respectively. Intercept and slope values,obtained by linear regression of the kobs versus inhibitor concentration plot (eqn 2), yielded the association and

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society 3753

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

dissociation rate constants kon and koff, respectively, and the inhibition constant (eqn 3)

KappI ¼ koff =k

appon (3)

Kappon was determined from the slope of the plot and then corrected for substrate competition using the

equation

ktrueon ¼ kappon 1þ [S]Km

� �(4)

The KtrueI value was obtained from the equation

KtrueI ¼ koff =k

trueon (5)

Crystallization and data collectionCrystallization of the AGT–DCS complex was attempted by co-crystallization using the vapor diffusion(hanging drop) technique: 1 ml of protein/ligand solution (215 mM AGT; 2 mM DCS; KP 50 mM pH 7.4) wasmixed with an equal volume of reservoir: 10–12% PEG 6K; 5% 2-methyl-2,4-pentanediol (MPD); 0.1 M MESpH 6.5. Single crystals were cryoprotected by fast soaking into a reservoir solution containing 2 mM DCS and25% MPD. For the AGT–LCS complex, AGT native crystals grown in the described conditions, with the excep-tion of DCS, were soaked into a solution containing the reservoir and 20 mM LCS and 20% MPD, before flashfreezing into N2(l) after 20 s. Complete data sets were collected at the ESRF in Grenoble on the ID23-1 beam-line (operated by the MxCuBE [27] software) at 2.7 and 3.0 Å resolution for the AGT–DCS and AGT–LCScrystals, respectively. Data were indexed and integrated with XDS [28] and scaled with Aimless [29] within theCCP4 software suite [30]. Both crystals belonged to the P41212 space group with one monomer in the asym-metric unit. Initial phases were obtained by molecular replacement using MOLREP [31] and 1.7 Å resolutionstructure of AGT as template (pdb: 5F9S [8]). Cycles of model building and refinement were performed withCOOT [32] and REFMAC5 [33], respectively. The models geometry was validated with MolProbity [34] beforedeposition in the protein data bank with accession code 6RV0 and 6RV1 for the AGT–DCS co-crystallizationand for the LCS soaked crystal, respectively. Data collection and refinement statistics are reported inSupplementary Table S1.

UPLC analysesAn amount of 10 mM AGT was inactivated in presence of 2 mM LCS or DCS at 25°C. After 2 h, the reactionswere stopped by the addition of TCA 10% (v/v) and the samples were centrifuged at 18 000g for 20 min fol-lowed by filtration using Millex-GV 0.22 mm filters (Merck MilliPore Ltd). Aliquots were analyzed by UPLC(Waters UPLC Acquity H class) using a Waters Cortecx C18+, 2.7 mm, 2.1 mm × 150 column equilibrated with50 mM potassium phosphate pH 2.35 [35]. An isocratic elution was used for 6 min. Each injection was 3 mlwith a flow rate of 0.2 ml/min and absorbance was monitored at 295 nm.

Mass spectrometry (MS)To identify small molecule reaction products, AGT was inactivated with LCS or DCS as described for UPLCanalyses, and each sample was deproteinized by TCA 10% (v/v). High-resolution mass spectra were recordedon a Thermo Scientific LTQ Orbitrap electrospray ionization (ESI) ion trap mass spectrometer. Samplevolumes were typically 5 ml under automated injection into a LC system equipped with a C18 column followedby separation in a methanol : water (50 : 50) mobile phase. The mass spectrometer was set up in full scanmode, with a range from 150 to 2000 m/z and a resolution of 60 000.To analyze the LCS- or DCS-inactivated AGT by MS, 10 mg of each protein sample were diluted in 100 ml

of 50 mM ammonium bicarbonate, reduced with 10 mM dithiothreitol (Wako Co.) and then alkylated with55 mM iodoacetamide (Wako Co.). Overnight digestion was done with Lys-C/Trypsin combo (Promega Co.) ina 1 : 50 enzyme : protein ratio. Upon stopping the digestion with 1% trifluoroacetic acid, the peptide mixture

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society3754

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

was cleaned with desalting C18 tips (StageTip, Thermo), vacuum-centrifuged to dryness, and re-suspendedwith 30 ml of 0.1% formic acid in water for LC/MS analysis. Data were collected using a Q-Exactive Plus massspectrometer (ThermoFisher Scientific) coupled with Dionex Ultimate 3000RS liquid chromatography system(ThermoFisher Scientific). An analytical column Zorbax 300SC-18 0.3 × 150 mm, 3.5 mm (Agilent Co.) wasused for sample chromatographic separation. Peptides were separated at a flow rate of 3.5 ml/min using a gradi-ent of 1–42% ACN (0.1% formic acid) over 60 min.Raw data files were searched against a composite target/decoy database using SEQUEST HT from Proteome

Discoverer software (PD, v.2.2, Thermo Fisher Scientific). The E. coli protein database was downloaded fromUniProt (UP000000625), combined with contaminants database (cRAP, https://www.thegpm.org/crap) andtarget protein sequence AGXT (Uniprot accession P21549). MS2 spectra were searched with ±20 ppm for pre-cursor ion mass tolerance, ±0.1 Da for fragment ion mass tolerance, using trypsin as enzyme, two missed clea-vages maximum, dynamic modifications for oxidation of methionine and deamidation of glutamine andasparagine, fixed modification carbamidomethylation of cysteine residues and variable modification of lysinewith PLP (mass shift +231.02911, C8H10NO5P). Label-free quantification based on peak intensity, usingunique and razor peptides, and normalized using total protein amount. Extracted ion chromatogram (XIC) andisotopic ion distribution were done using Xcalibur software (ThermoFisher Scientific).

Docking studiesThe crystal structure of human AGT (PDB ID: 1H0C) (7) was protonated using the Protonate 3D tool of MOE2018.0101 and the docking site was selected using the MOE Site Finder tool. LCS and DCS structures weredocked into the active site of wild-type AGT using the MOE Dock tool. As a first step, Placement generates acollection of ligand conformations using the bond rotation method. These are then placed in the site by theTriangle Matcher method and ranked using the London dG scoring function, which estimates the free energyof binding of the ligand from a given pose. Finally, the Refinement tool is used for energy minimization of thepocket, before rescoring the poses with Affinity dG, which estimates the enthalpic contribution to the freeenergy of binding. The top 30 poses for each ligand were output in the MOE database and analyzed by visualinspection of the structure.

Cell culture and lysisChinese hamster ovary cells stably expressing glycolate oxidase (CHO-GO), and CHO-GO clones expressingwild-type AGT (CHO-GO-AGT-wt) or the G41R variant (CHO-GO-AGT-G41R), were cultured in Ham’s F12Glutamax medium supplemented with fetal bovine serum (10%, v/v), penicillin (100 units/ml) and strepto-mycin (100 mg/ml) at 37°C in a 5% CO2 humidified environment. The expression of AGT and GO was main-tained by adding G-418 (0.8 mg/ml) and Zeocin (0.4 mg/ml), respectively, to the culture medium. To test theeffects of LCS or DCS, cells were cultured for 7 days in the absence or presence of one of the two CS enantio-mers at 100 mM concentration. At the end of treatment, cells were harvested and lysed by freeze/thawing (fivecycles) in Phosphate Buffer Saline (PBS), pH 7.2 supplemented with 100 mM PLP and protease inhibitor cock-tail (Complete Mini, Roche). The whole-cell extract was separated by centrifugation (29 200g, 10 min, 4°C) toobtain the soluble fraction. Protein concentration in the soluble fraction was measured using the Bradfordprotein assay.

Western blot analysesAn amount of 5 mg of soluble cell lysate were loaded on 10% SDS–PAGE, transferred on a nitrocellulose mem-brane and immunoblotted with anti AGT antibody (1 : 10 000) in 2.5% (w/v) milk in TTBS (50 mM Tris–HCl,pH 7.5, 150 mM NaCl, 0.1% Tween 20) overnight at 4°C. After three washes in TTBS, the membrane was incu-bated with peroxidase-conjugated antirabbit immunoglobulin G (IgG) (1 : 10 000) in 5% milk in TTBS for 1 hat room temperature. Immunocomplexes were visualized by an enhanced chemiluminescence kit (ECL, PierceBiotechnology, Rockford, IL).

Glycolate toxicity assayCHO-GO, CHO-GO-AGT-wt, and CHO-GO-AGT-G41R cells were seeded in Petri dishes in Ham’s F12Glutamax medium. Where indicated, 100 mM DCS was added to the culture medium. After 6 days cells weredetached and seeded in 96-well plates at a density of 8000 cells/well and exposed to 0.5 mM glycolate. After

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society 3755

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

24 h of treatment cell viability was assessed by crystal violet staining and results were normalized to that ofcontrols grown without glycolate, as previously described [36].

Statistical analysisExperiments have been performed at least in triplicate and in any case the standard error mean (SEM) was<10%. Statistical analyses were performed by Origin® 7 (Origin Lab) or GraphPad Prism Version 6.0(GraphPad software, San Diego, CA, USA).

ResultsBinding of LCS to wild-type and mutant AGTWe evaluated the interaction of LCS with wild-type AGT (encoded by the most common major allele) [37] andthe G41R variant (on the background of the major allele) by absorbance spectroscopy. As previously reported,purified holoAGT displays an absorbance peak at 423 nm and a shoulder at ∼340 nm corresponding to theoxoenamine and enolimine tautomers of the internal aldimine, respectively [1]. The addition of 100 mM LCS

Figure 1. Spectral changes and fluorescence spectra of AGT and the G41R variant upon addition of LCS.

(A) Time-dependent absorbance spectra of 9 mM wild-type AGT in the absence (black line) or in the presence of 100 mM LCS (dotted line) at the

indicated times. (B) Time-dependent absorbance spectra of 9 mM G41R-Ma variant in the absence (black line) or in the presence of 100 mM LCS

(dotted line) at the indicated times. (C) Fluorescence emission intensity upon excitation at 380 nm for wild-type AGT (—) and the G41R variant (‐ ‐ ‐)

at 1 mM protein concentration. (D) Fluorescence emission intensity upon excitation at 334 nm for wild-type AGT (—) and at 325 nm for the G41R

variant (—) at 1 mM protein concentration. The experiments have been performed in KP 0.1 M pH 7.4 at 25°C.

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society3756

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

to wild-type AGT caused the disappearance of the band at 423 nm and the concomitant appearance of a peakat 334 nm, with a shoulder at 380 nm (Figure 1A). The same spectroscopic changes occurred upon LCS add-ition to the G41R variant, although in the mutant the band was centered at 325 nm (Figure 1B). Upon excita-tion at 380 nm, the AGT–LCS and the G41R–LCS complexes showed a fluorescence emission spectrum withmaximum at 450 nm, typical of the oxime formed between the ligand and PLP [38] (Figure 1C). On the otherhand, the bands a 334 nm and at 325 nm produced a fluorescence emission at 408 nm and 416 nm for wild-type and mutant AGT, respectively (Figure 1D). These features are indicative of coenzyme forms with a C40

showing sp3 hybridization [39]. Based on the known mechanisms of interaction of CS with PLP-enzymes, dif-ferent hypotheses on the attribution of the 334 and 325 nm bands can be formulated. One of them is the for-mation of an inactive hydroxyisoxazole-PMP adduct, as previously reported for alanine racemase andbranched-chain aminotransferase [14,15]. However, we found that the addition of the glyoxylate ketoacid to theenzyme–LCS complex led to the recovery of the ketoenamine tautomer of the internal aldimine(Supplementary Figure S1A). This suggests that the band is related to an active form of the coenzyme, such asPMP or a ketimine intermediate, which undergoes reverse half-transamination in the presence of an aminoacceptor [5,22].

Figure 2. Crystallographic analysis of AGT interaction with LCS and DCS.

(A) Omit map showing the positive electron density (Fo–Fc map: in green contour level 3.0 sigma — 0.14 e/Å3) of the AGT

crystal soaked with LCS. (B) Electron density maps after refinement of the final model including PLP internal aldimine (purple)

with an occupancy of 0.75; 2Fo–Fc map (contour 1.5 sigma) in blue. A small positive peak in the Fo–Fc map (in green contour

level 3.0 sigma) is also observed near the PLP C40 atom suggesting that a small amount of PMP may also be present in the

active site. (C) Omit map showing the positive electron density (Fo–Fc map: in green contour level 3.0 sigma — 0.15 e/Å3) of

the AGT–DCS co-crystallization crystal. Maps were calculated omitting K209, and PLP from the model. (D) Electron density

maps after refinement of the final model including PMP (pink) with an occupancy of 0.75; 2Fo–Fc map (contour 1.5 sigma) in blue.

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society 3757

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

To gain structural details of the AGT–LCS complex, we crystallized wild-type holoAGT and we soaked thecrystals into a solution containing LCS. We collected data at 3.0 Å resolution (Supplementary Table S1). In theomit map (Figure 2A) we observed a continuous electron density for the Schiff base bond between PLP andLys209. After refinement of a model including the internal aldimine, the presence of a weak positive peak nearthe C40 of PLP and the low density of the Lys209 side chain suggested the presence of a mixed situation. Itmust be clarified that the electron density maps represent the average state of the active site among all the pro-teins in the crystal. In this case the data suggest that, although the majority of the proteins display the internalaldimine at the active site, a small percentage of PMP may also be present in the crystal (Figure 2B). However,given the low resolution of these data, it was not possible to refine neither PMP nor a mixed model with boththe internal aldimine and PMP at different occupancies in the active site, and the final model consisted of theinternal aldimine with an occupancy of 0.75 (i.e. 75% of all the proteins of the crystal).

LCS inhibition mechanismLCS led to a concentration-dependent drop in transaminase activity, confirming its role as inhibitor, with IC50

values of 50 mM and 136 mM for wild-type and G41R mutant AGT, respectively. Importantly, LCS inhibitionwas reversible, since the addition of L-alanine at saturating concentration to a preformed AGT–LCS complexcaused an almost complete recovery (90%) of catalytic activity for both wild-type and variant. The IC50 valueof both species did not change as a function of the incubation time (Supplementary Table S2), thus excludingthe occurrence of a time dependent inhibition. Since LCS undergoes half-transamination, we hypothesized thatit could be a competitive inhibitor for AGT. Indeed, the kinetic parameters of the enzymes for the substrateL-alanine indicate that the presence of LCS did not affect the value of Vmax, while it caused a concentration-dependent increase in the KM value. By a global fitting of the data we calculated Ki values of 37 ± 4 mM and 61± 10 mM for wild-type AGT and the G41R variant, respectively (Figure 3). It should be noted that the Ki valueof the variant is 2-fold higher than that of the wild-type, in line with previous data showing that the Gly41mutation induces slight changes of the AGT active site architecture [22].

Binding of DCS to wild-type and mutant AGTThe addition of 1 mM DCS to wild-type AGT caused a time-dependent disappearance of the absorbance bandsat 420 and 340 nm and the concomitant appearance of an absorbance band at 370 nm (Figure 4A). The sameband also appeared upon DCS addition to the G41R variant, although in this case an absorbance band at331 nm was also present, suggesting the formation of two species (Figure 4B). The excitation at 370 nm of thetwo enzyme–DCS complexes generated a fluorescence emission spectrum with maximum at 450 nm, indicatingthe presence of an oxime intermediate (Figure 4C) [13]. On the other hand, the band at 331 nm of the variantcould be ascribed either to the inactive enolimine tautomer of the internal aldimine, or to the formation of aketimine intermediate.

Figure 3. Determination of the kinetic parameters of wild-type AGT and the G41R variant for L-alanine in presence of LCS.

(A) Cumulative fitting of kinetic parameters of wild-type AGT for L-alanine in absence or presence of LCS at the indicated

micromolar concentrations (eqn 3). (B) Cumulative fitting of kinetic parameters of the G41R variant vs L-alanine in absence or

presence of LCS at the indicated micromolar concentrations (eqn 3). Experiments have been performed in KP 0.1 M pH 7.4 at 25°C.

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society3758

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

We co-crystallized wild-type AGT in the presence of DCS at 2.5 Å resolution (Supplementary Table S2).Inspection of the electron density maps in the active site suggested the absence of the Schiff base bond betweenLys209 and the C40 of PLP (Figure 2C). Additional electron density on the opposite side of the C40 atomfurther suggested that the internal aldimine had reacted with DCS. After several attempts the best fit for theobserved electron density was PMP, which refined with reasonable B-factors and real space correlation coeffi-cients (RSCC) values [38,40] when included in the model with a 0.75 occupancy (Figure 2D). Also in this case,from the density and B-factor analysis the presence of unreacted internal aldimine and/or apo active sites maynot be excluded.

Inhibition mechanism mediated by DCSSince the absorbance spectra evidenced a time dependent binding of DCS, we determined the IC50 upon incu-bation of the wild-type enzyme and the G41R variant with the inhibitor at different time points (20 min,60 min and overnight) at 25°C. The values reported in Supplementary Table S2 show a decreasing IC50 value atincreasing incubation time, a behavior consistent with time-dependent inhibition. Nevertheless, the addition of

Figure 4. Spectral changes and fluorescence emission spectra of wild-type AGT and the G41R variant upon addition of DCS.

(A) Time-dependent absorbance spectra of 9 mM wild-type AGT in the absence (black line) or in the presence of 1 mM DCS (dotted lines) at the

indicated times. (B) Time-dependent absorbance spectra of 9 mM G41R in the absence (black line) or in the presence of 5 mM DCS (dotted lines) at

the indicated times. (C) Fluorescence emission spectra of 1 mM wild-type AGT (—) and of the G41R variant (‐ ‐ ‐) in complex with DCS upon

excitation at 370 nm. Experiments have been performed in KP 0.1 M pH 7.4 at 25°C.

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society 3759

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

L-alanine at saturating concentration to a preformed AGT–DCS complex caused a time-dependent and almostcomplete recovery of catalytic activity, while the incubation of the wild-type AGT–DCS and G41R–DCS com-plexes with PLP alone led to a recovery of activity of 23% and 54%, respectively (Supplementary Figure S2).These data indicate that DCS inhibition can be completely reversed by the substrate but not by the coenzyme,

Figure 5. Time dependent inhibition of wild-type AGT and the G41R variant by DCS.

(A) Progression curves of wild-type AGT (0.1 mM) in the presence of the indicated concentrations of DCS in KP 0.1 M pH 7.4 at

25°C. (B) Trends of the kobs values obtained using equation 4 as a function of inhibitor concentration. (C) Progression curves of

the G41R variant (0.2 mM) in the presence of the indicated concentrations of DCS in KP 0.1 M pH 7.4 at 25°C. (D) Trends of

the kobs values obtained using equation 4 as a function of inhibitor concentration. (E) Trends of the kobs values for wild-type

AGT (closed symbols) and G41R (open symbols) obtained at 0.1 mM enzyme concentration in the presence of 5 mM DCS, at

increasing substrate concentration (30–500 mM).

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society3760

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

thus suggesting that the ligand does not form an irreversibly modified coenzymatic form as reported for otherenzymes of the same family [14,15].To better investigate the mechanism underlying the time-dependent inhibition mediated by DCS, we deter-

mined the progression curves of the loss of enzymatic activity as a function of time in the presence of theinhibitor at increasing concentrations (Figure 5A–C). For each curve we calculated the kobs using equation 4. Itshould be considered that there are two common mechanisms of slow-binding inhibition. The first, called‘one-step slow binding inhibition’, is described by the scheme: E + I↔ EI and involves the direct formation ofan inactive enzyme-inhibitor complex. The second, called ‘two-step slow binding inhibition’ or ‘isomerizationslow binding inhibition’, is described by the scheme: E + I↔ EI↔ EI* and involves the formation of theenzyme–inhibitor complex followed by an isomerization generating an inactive complex. To distinguishbetween the two models, we plotted the apparent first-order rate constant (kobs) against inhibitor concentration.As reported in Figure 5B,D, for both wild-type AGT and the G41R variant we observed a linear trend, indicat-ing that the reaction kinetics follows a one-step slow-binding inhibition mechanism. From the linear fit of thedata to equation 5, we calculated the association and dissociation rate constants, and then the apparentenzyme–DCS equilibrium dissociation constants (KI

app). To define the modality by which the inhibitor interactswith the enzyme, we determined the kobs values from progression curves obtained at a fixed saturating concen-tration of DCS in the presence of increasing concentrations of the substrate. As shown in Figure 5E, the kobsvalues decrease as substrate concentration increases. This trend is indicative of a competitive inhibition andallows to calculate the true KI values from the KI

app. The kinetic constants describing DCS inhibition are sum-marized in Table 1.

Identification of the wild-type and mutant AGT-cycloserine reaction productsTo exclude the occurrence of covalent modifications upon interaction with CS, we performed MS analyses afterlimited proteolysis of AGT samples before and after inactivation by LCS or DCS. As expected, the results con-firmed the presence of a covalent PLP–Lys209 complex in the untreated sample. The same complex was alsopresent in the AGT–LCS and AGT–DCS sample, although its relative abundance was 7% and 29%, respectively,as compared with untreated AGT (Supplementary Figure S3), as expected considering that the internal aldi-mine is converted to other reaction intermediates upon reaction with CS (see below). Thus, AGT inhibitionwas not due to the covalent modification of any active site residue.We then performed UPLC analyses to determine the changes in coenzyme content using a previously setup

method [35]. As reported in Figure 6A, upon reaction with LCS 77% of the initial PLP content of AGT is con-verted to PMP, in agreement with what observed in the presence of L-alanine and confirming the occurrence ofan half-transamination. On the other hand, upon reaction with DCS only 4% of PLP was converted to PMP,while 24% eluted as unmodified PLP and the remaining portion eluted as unidentified small peaks appearing atelution volumes higher than that of PLP. Notably, upon addition of L-alanine to the AGT–DCS complex, 80%of the coenzyme is converted to PMP. This further confirms that DCS forms a reversible complex with AGT,which can be displaced by the substrate. We then analyzed the AGT–DCS reaction products by ESI–MS(Figure 6B). We found (i) peaks at m/z 330.77 and 332.81, which can be attributed to the PLP–DCS externalaldimine, or to a ketimine intermediate generated by a 1,3-prototropic shift, and (ii) peaks at m/z 348.78 and350.78, which are compatible with the addition of a water molecule to the external aldimine complex leading toopening of the CS ring. The latter peaks could be ascribed to the external aldimine of PLP with formedβ-aminoserine, an oxime generated by intramolecular rearrangement, or a ketimine between PMP andβ-aminooxypyruvate [17,41]. Our spectroscopic and crystallographic data, along with the reversibility of theAGT inhibition process, allow to exclude the formation of a covalent PMP-isoxazole complex [14]. Moreover,no PMP formation is seen by UPLC, thus excluding the occurrence of an half-transamination. Thus, we canattribute the peaks at m/z 348.78 and 350.78 to an oxime intermediate.

Table 1. Kinetic constants of the time-dependent inhibition of DCSfor wild-type AGT and the G41R variant

Enzyme kon (mM−1 min−1) koff (min−1) KItrue (mM)

Wild-type 9.1 ± 0.4 × 10−2 1.6 ± 0.2 × 10−2 175 ± 23

G41R 4.1 ± 0.2 × 10−2 2.8 ± 0.5 × 10−2 682 ± 134

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society 3761

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

We obtained a qualitatively similar ESI–MS pattern by analyzing the AGT–LCS reaction mixture, althoughin this case the intensity of the peaks at m/z 348.78 and 350.78 was lower than that obtained in the AGT–DCSsample (data not shown). It should be noticed that the obtained m/z values differ by 0.7 D from the expectedvalues of the nominal mass of the PLP-CS complexes, probably due to already reported random errors thataffect accuracy giving bias in mass measurements [42].

Effects of CS enantiomers on specific activity and expression level ofwild-type AGT and the G41R variant in CHO-GO cellsTo test if LCS and/or DCS could act as PCs for AGT variants showing folding defects, we took advantage of acellular model made up of CHO cells that overexpress glycolate oxidase (CHO-GO) and recapitulate hepaticglyoxylate metabolism [43]. In the presence of glycolate in the culture medium, these cells produce glyoxylateinside peroxisomes and show a strongly reduced viability. CHO-GO cells stably transformed with wild-typeAGT (CHO-GO-AGT-wt) or the G41R variant (CHO-GO-AGT-G41R) were grown in the presence or in the

Figure 6. UPLC and ESI–MS analysis of AGT–CS reaction products.

(A) Coenzyme content of wild-type AGT upon 2 h reaction with 2 mM LCS or DCS at 25°C, as obtained by UPLC analysis. The reaction of the

enzyme with saturating L-alanine was used as positive control. (B) ESI–MS analysis of the wild-type AGT–DCS reaction mixture. The chemical

structures attributed to each peak are shown. See the text for details.

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society3762

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

Figure 7. Effects of LCS or DCS treatment on specific activity and expression level of AGT. Part 1 of 2

CHO-GO-AGT-wt and CHO-GO-AGT-G41R cells were grown for 7 days in the presence or in the absence of 100 mM LCS or

DCS. At the end of treatment cells were detached, lysed and soluble fraction of each sample used for experiments: (A) AGT

enzymatic activity determination. Specific activity of AGT in CHO-GO-AGT-wt control cells (192.54 ± 12.68 nmoles of pyruvate/

min/mg protein) was assumed as 100%. Data represent mean ± SEM (n = 4). * P < 0.005, Student’s t-test; (B) AGT expression

level quantification. AGT levels in CHO-GO-AGT-wt control cells was assumed as 100%. (C) Representative western blot of

AGT expression levels in the indicated cells clones. For each gel, the corresponding Ponceau staining of the total lysate is also

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society 3763

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

absence of 100 mM LCS or DCS in the culture medium for 7 days. CS concentration was chosen taking intoaccount the inhibition constant values, and the use of a pharmacologically relevant ligand concentration, whilethe treatment time was that sufficient to allow protein synthesis and folding in the presence of the treatingagent, considering the AGT half-life [44]. Upon cell harvesting and lysis, the soluble fraction of the lysate wasused for experiments aimed at measuring the AGT expression level and specific activity. We found that neitherLCS nor DCS alter the transaminase specific activity and the AGT expression level of CHO-GO-AGT-wt cells(Figure 7). As for the G41R variant, data obtained with the purified protein indicate that it displays an enzym-atic activity corresponding to ∼10% that of wild-type AGT (in terms of kcat). When expressed in CHO-GOcells, the variant showed 10% specific activity and 8% expression level in the soluble fraction, as compared withthe wild type. These data confirm that the variant is endowed with a folding defect that reduces the amount ofsoluble and active protein present in the cell, in line with previous evidences indicating a higher tendency toaggregation induced by the Gly41 mutation [22]. Notably, LCS did not affect the specific activity and theexpression level of the variant. Since LCS undergoes half-transamination, we also tested the chaperone effectupon replenishment of LCS in the medium every 24 h. However, we did not notice any change in AGT expres-sion or activity. On the other hand, DCS induced a statistically significant increase in specific activity(Figure 7A). To understand whether DCS-induced enhancement of enzymatic activity was dependent on anincreased expression, we analyzed the protein levels CHO-GO-AGT-G41R cells (Figure 7B) and we found thatDCS treatment mildly increases the amount of AGT present in the soluble fraction. These data suggest that theincreased AGT specific activity of CHO-GO-AGT-G41R cells treated with DCS might depend on the capacityof this enantiomer to specifically bind the misfolded variant and improve its folding efficiency. In agreementwith these results, the treatment with DCS significantly improved the glyoxylate detoxification ability ofCHO-GO cells expressing the G41R variant (Figure 7C), because it increases cell viability upon glycolate treat-ment. These data confirm that the treatment increases the amount of intraperoxisomal functional AGT able tometabolize glyoxylate.

DiscussionThis work is focused on the analysis of the interaction of AGT, a clinically relevant enzyme involved in the raredisorder PH1, with the two enantiomers of CS. We performed a spectral, kinetic, and structural characteriza-tion of the interaction of LCS and DCS with the purified enzyme both in the wild-type form encoded by themajor allele and in a mutant form bearing a folding defect (G41R) on the background of the major allele [22].We also tested the possible action of LCS and DCS as PCs, i.e. their ability to counteract the effects of theG41R pathogenic mutation in a cellular model.Our results indicate that both LCS and DCS do not covalently modify AGT, but bind at the active site and

reversibly inhibit transaminase activity in a competitive way. However, they show a different binding kineticsand generate a different equilibrium of species. Spectral and kinetic studies agree with LCS behaving as a clas-sical competitive inhibitor that instantaneously undergoes transamination producing a ketimine and a smallfraction of an oxime intermediate. Crystallographic data are in agreement with the occurrence of a half-transamination, although we did not obtain other clues about the product(s) and intermediates of the reaction.Since the AGT crystals were soaked with LCS, it must be considered that the reactivity in the crystal form islower with respect to that in solution due to the slower diffusion of the ligand within the crystal lattice.Therefore, the reaction could have been incomplete, but longer soaking times resulted in non-diffracting crys-tals. Nevertheless, the UPLC analysis of the coenzyme content upon acidic deproteinization clearly confirmsthe formation of PMP, and ESI–MS data show a peak compatible with a ketimine intermediate.On the other hand, DCS shows a time-dependent binding to wild-type AGT and to the G41R mutant,

leading to spectral changes consistent with the formation of an oxime intermediate associated with the loss ofcatalytic activity. The kinetic analysis of the inhibition indicates that it follows a single-step competitive

Figure 7. Effects of LCS or DCS treatment on specific activity and expression level of AGT. Part 2 of 2

shown as loading control. (D) Indirect glycolate toxicity assay. Histogram representative of cell viability after 24 h of treatment

with 0.5 mM glycolate expressed as percentage with respect to untreated cells. Six replicates have been measured for each

sample. CHO-GO and CHO-GO-AGT-wt cells represent the negative and positive control, respectively. The images are

representative of one out of three separate experiments. Data represent mean ± SEM (n = 4).

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society3764

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

mechanism. It is not easy to explain why we did not detect the oxime intermediate in the crystal structure ofthe AGT–DCS complex, but we can speculate that it could undergo hydrolysis during the crystallizationprocess or even radiation damage, as previously reported for other PLP intermediates [8]. UPLC and ESI–MSdata confirm the reversibility of DCS binding and are consistent with the formation of an oxime generated bywater addition to the isoxazolidone ring, followed by intramolecular rearrangement.Overall, the mechanism that best recapitulates the results on the reaction of AGT with LCS or DCS is

reported in Figure 8. The enzyme in the PLP-form reacts with LCS or DCS forming an external aldimine. Inthe case of LCS, a ketimine is formed upon direct deprotonation at Cα and reprotonation at the coenzyme C40.However, it cannot be excluded that the 1,3-prototropic shift occurs on the external aldimine intermediateformed upon water addition to the CS ring. Molecular docking analyses of the AGT–LCS complex(Supplementary Figure S4) indicate that LCS is mainly held in place through a ring stacking interaction withTyr260 of the neighboring subunit, and by hydrogen bonds between the exocyclic oxygen and Arg360.Notably, Lys209 is at 3.5 Å from Cα, thus in a favorable position to perform the 1,3-prototropic shift leading toketimine formation. In the case of the AGT–DCS complex, the ligand moiety assumes a different conformation,with the amino group and the exocyclic oxygen interacting with Arg360. Thus, transaldimination is lessfavored, in agreement with the slow kinetics of DCS binding. Moreover, Lys209 is not properly positioned toact as acid/base catalyst, thus explaining why the AGT–DCS complex undergoes a ring-opening processmediated by water addition, a process probably favored by the liberation of the carboxylate group that can pref-erentially interact with Arg360. The latter step generates an intermediate that undergoes an intramolecularrearrangement producing an oxime of PLP with β-aminooxyserine [7].The mechanism here described also explains why we noticed that both LCS and DCS are reversible inhibitors

of AGT, as demonstrated by the recovery of activity in the presence of saturating concentrations of the physio-logical substrate L-alanine. LCS behaves as a classical competitive inhibitor, and thus can be displaced by thesubstrate when it is present at saturating concentration in the overall transamination reaction mixture. In add-ition, the ketimine formed with LCS is supposed to be in equilibrium with the external aldimine intermediate,

Figure 8. Proposed reaction mechanism of AGT with LCS or DCS.

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society 3765

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

thus explaining the recovery of activity when the AGT–LCS complex is preformed. A similar behavior has beenalready reported for competitive inhibitors of alpha-galactosidase A in Fabry disease [45,46] as well as for imi-nosugar derivatives as glucocerebrosidase inhibitors in Gaucher disease [47,48]. As for the oxime intermediategenerated with DCS, it is not difficult to hypothesize that the α-amino group of L-alanine could compete withthe N atom of ligand moiety for the binding to the C40 of PLP and displace the inhibitor, thus leading to arecovery of catalytic activity. These findings get a new light on the mechanism underlying the action of LCSand DCS, because both molecules have been always considered as irreversible inhibitors of PLP-enzymes[15,49]. However, it should be pointed out that experiments aimed at testing the regain of activity induced bythe substrate have not been performed in all previous studies.The reversible behavior of LCS and DCS prompted us to test their action as PCs in a cellular model recapitu-

lating the metabolic alterations typical of patients affected by PH1, a disease where misfolding is one of themain factors underlying the effects of pathogenic mutations in AGT. We used as models cells expressing wild-type AGT and cells expressing the G41R variant, which shows a folding defect giving rise to an increased ten-dency to aggregation and to intracellular degradation [22,23]. We found that only DCS is able to partly rescuefor the pathogenic G41R mutation, leading to a recovery of catalytic activity paralleled by an increased portionof protein present in the soluble fraction. This can be understood considering that LCS behaves as a substrate,generating PMP, which can rapidly dissociate from the G41R variant inside the cell. On the other hand, theslow binding of DCS could allow the ligand to remain bound at the active site for a time long enough topromote the correct folding of the protein. It should be mentioned that DCS is a natural product used to treatresistant Mycobacterium tuberculosis infections as well as neurological disorders since it is a potent NMDAreceptor agonist [50]. These findings strongly support the use of DCS as a valuable template for the synthesisof PCs for PH1. The preparation and biological evaluation of improved analogs is currently ongoing in ourlaboratories.

Abbreviations2,4-DNP, 2,4-dinitrophenylhydrazine; AGT, alanine:glyoxylate aminotransferase; CS, cycloserine; DCS,D-cycloserine; ESI, electrospray ionization; KP, potassium phosphate buffer; LCS, L-cycloserine; LDH, L-lactatedehydrogenase; MPD, 2-methyl-2,4-pentanediol; MS, mass spectrometry; PBS, phosphate-buffered saline; PC,pharmacological chaperone; PH1, primary hyperoxaluria type 1; PLP, pyridoxal 50-phosphate; PMP,pyridoxamine 50-phosphate.

Authors contributionM.D., S.G., G.C., M.P., B.C.: conceived and designed the experiments. M.D., S.G., G.A., G.G., G.P., A.F.:performed the experiments and analyzed the data. M.P., G.G., F.C., G.C., P.L.: analyzed the data and providedsupport for mass spectrometry and crystallography experiments. M.D., S.G., B.C. wrote the paper. All authorsproof read and approved the manuscript.

FundingThis work has been supported by a grant from the Oxalosis and Hyperoxaluria Foundation [OHF2016] to B.C.,and by grants from Sapienza University of Rome, Italy [RP11715C644A5CCE and RG11816430AF48E1] to F.C.

AcknowledgementsWe thank the staff of ESRF and EMBL-Grenoble for assistance and support in using beamline ID23-1 and Dr.Alejandro Villar-Briones from the IAS in OIST for LC–MS experiments. The Centro Interdipartimentale Misure‘‘G. Casnati’’ is kindly acknowledged for the contribution in the analytical determination of the moleculesreported in this study. M.D. thanks Japan Society for the Promotion of Science (JSPS). Fellowship number:P19764

Competing InterestsThe authors declare that there are no competing interests associated with the manuscript.

References1 Cellini, B., Bertoldi, M., Montioli, R., Paiardini, A. and Borri Voltattorni, C. (2007) Human wild-type alanine:glyoxylate aminotransferase and its naturally

occurring G82E variant: functional properties and physiological implications. Biochem. J. 408, 39–50 https://doi.org/10.1042/BJ20070637

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society3766

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

2 Cochat, P. and Rumsby, G. (2013) Primary hyperoxaluria. N. Engl. J. Med. 369, 649–658 https://doi.org/10.1056/NEJMra13015643 Hoppe, B. (2012) An update on primary hyperoxaluria. Nat. Rev. Nephrol. 8, 467–475 https://doi.org/10.1038/nrneph.2012.1134 Cochat, P., Fargue, S. and Harambat, J. (2010) Primary hyperoxaluria type 1: strategy for organ transplantation. Curr. Opin. organ Transplant. 15,

590–593 https://doi.org/10.1097/MOT.0b013e32833e35f55 Dindo, M., Conter, C., Oppici, E., Ceccarelli, V., Marinucci, L. and Cellini, B. (2019) Molecular basis of primary hyperoxaluria: clues to innovative

treatments. Urolithiasis 47, 67–78 https://doi.org/10.1007/s00240-018-1089-z6 Hoyer-Kuhn, H., Kohbrok, S., Volland, R., Franklin, J., Hero, B., Beck, B.B. et al. (2014) Vitamin B6 in primary hyperoxaluria I: first prospective trial after

40 years of practice. Clin. J. Am. Soc. Nephrol. 9, 468–477 https://doi.org/10.2215/CJN.068206137 Zhang, X., Roe, S.M., Hou, Y., Bartlam, M., Rao, Z., Pearl, L.H. et al. (2003) Crystal structure of alanine:glyoxylate aminotransferase and the relationship

between genotype and enzymatic phenotype in primary hyperoxaluria type 1. J. Mol. Biol. 331, 643–652 https://doi.org/10.1016/S0022-2836(03)00791-5

8 Giardina, G., Paiardini, A., Montioli, R., Cellini, B., Voltattorni, C.B. and Cutruzzola, F. (2017) Radiation damage at the active site of human alanine:glyoxylate aminotransferase reveals that the cofactor position is finely tuned during catalysis. Sci. Rep. 7, 11704 https://doi.org/10.1038/s41598-017-11948-w

9 Mandrile, G., van Woerden, C.S., Berchialla, P., Beck, B.B., Acquaviva Bourdain, C., Hulton, S.A. et al. (2014) Data from a large European studyindicate that the outcome of primary hyperoxaluria type 1 correlates with the AGXT mutation type. Kidney Int. 86, 1197–1204 https://doi.org/10.1038/ki.2014.222

10 Oppici, E., Montioli, R. and Cellini, B. (2015) Liver peroxisomal alanine:glyoxylate aminotransferase and the effects of mutations associated with primaryhyperoxaluria type I: an overview. Biochim. Biophys. Acta 1854, 1212–1219 https://doi.org/10.1016/j.bbapap.2014.12.029

11 Pey, A.L., Albert, A. and Salido, E. (2013) Protein homeostasis defects of alanine-glyoxylate aminotransferase: new therapeutic strategies in primaryhyperoxaluria type I. BioMed. Res. Int. 2013, 687658 https://doi.org/10.1155/2013/687658

12 Leidenheimer, N.J. (2018) Pharmacological chaperones: beyond conformational disorders. Handb. Exp. Pharmacol. 245, 135–153 https://doi.org/10.1007/164_2017_68

13 Oppici, E., Montioli, R., Dindo, M., Maccari, L., Porcari, V., Lorenzetto, A. et al. (2015) The chaperoning activity of amino-oxyacetic acid onfolding-defective variants of human alanine:glyoxylate aminotransferase causing primary hyperoxaluria type I. ACS Chem. Biol. 10, 2227–2236https://doi.org/10.1021/acschembio.5b00480

14 Amorim Franco, T.M., Favrot, L., Vergnolle, O. and Blanchard, J.S. (2017) Mechanism-based inhibition of the Mycobacterium tuberculosisbranched-chain aminotransferase by d- and l-cycloserine. ACS Chem. Biol. 12, 1235–1244 https://doi.org/10.1021/acschembio.7b00142

15 Fenn, T.D., Stamper, G.F., Morollo, A.A. and Ringe, D. (2003) A side reaction of alanine racemase: transamination of cycloserine. Biochemistry 42,5775–5783 https://doi.org/10.1021/bi027022d

16 Gregory, T., Olson, M.F., Lau, S., Rinehart, K.L. and Silverman, R.B. (1998) An aromatization mechanism of inactivation of γ-aminobutyric acidaminotransferase for the antibiotic L-cycloserine. J. Am. Chem. Soc. 120, 12 https://doi.org/10.1021/ja971047w

17 Lowther, J., Yard, B.A., Johnson, K.A., Carter, L.G., Bhat, V.T., Raman, M.C. et al. (2010) Inhibition of the PLP-dependent enzyme serinepalmitoyltransferase by cycloserine: evidence for a novel decarboxylative mechanism of inactivation. Mol. Biosyst. 6, 1682–1693 https://doi.org/10.1039/c003743e

18 Prosser, G.A. and de Carvalho, L.P. (2013) Reinterpreting the mechanism of inhibition of Mycobacterium tuberculosis D-alanine:D-alanine ligase byD-cycloserine. Biochemistry 52, 7145–7149 https://doi.org/10.1021/bi400839f

19 Ringe DPMCWVOMM. (1998) d-cycloserine inactivation of d-amino acid aminotransferase leads to a stable noncovalent protein complex with anaromatic cycloserine-PLP derivative. J. Am. Chem. Soc 120, 7 https://doi.org/10.1021/ja972993f

20 Goff, D.C. (2017) D-cycloserine in schizophrenia: new strategies for improving clinical outcomes by enhancing plasticity. Curr. Neuropharmacol. 15,21–34 https://doi.org/10.2174/1570159X14666160225154812

21 Williams, R.D., Sgoutas, D.S., Zaatari, G.S. and Santoianni, R.A. (1987) Inhibition of serine palmitoyltransferase activity in rabbit aorta by L-cycloserine.J. Lipid Res. 28, 1478–1481 PMID:3430071

22 Cellini, B., Montioli, R., Paiardini, A., Lorenzetto, A., Maset, F., Bellini, T. et al. (2010) Molecular defects of the glycine 41 variants of alanine glyoxylateaminotransferase associated with primary hyperoxaluria type I. Proc. Natl Acad. Sci. U.S.A. 107, 2896–2901 https://doi.org/10.1073/pnas.0908565107

23 Fargue, S., Lewin, J., Rumsby, G. and Danpure, C.J. (2013) Four of the most common mutations in primary hyperoxaluria type 1 unmask the crypticmitochondrial targeting sequence of alanine:glyoxylate aminotransferase encoded by the polymorphic minor allele. J. Biol. Chem. 288, 2475–2484https://doi.org/10.1074/jbc.M112.432617

24 Cellini, B., Montioli, R., Bianconi, S., Lopez-Alonso, J.P. and Voltattorni, C.B. (2008) Construction, purification and characterization of untagged humanliver alanine-glyoxylate aminotransferase expressed in Escherichia coli. Protein Pept. Lett. 15, 153–159 https://doi.org/10.2174/092986608783489580

25 Cellini, B., Lorenzetto, A., Montioli, R., Oppici, E. and Voltattorni, C.B. (2010) Human liver peroxisomal alanine:glyoxylate aminotransferase: differentstability under chemical stress of the major allele, the minor allele, and its pathogenic G170R variant. Biochimie 92, 1801–1811 https://doi.org/10.1016/j.biochi.2010.08.005

26 Burt, H.J., Pertinez, H., Sall, C., Collins, C., Hyland, R., Houston, J.B. et al. (2012) Progress curve mechanistic modeling approach for assessingtime-dependent inhibition of CYP3A4. Drug Metab. Dispos. 40, 1658–1667 https://doi.org/10.1124/dmd.112.046078

27 Gabadinho, J., Beteva, A., Guijarro, M., Rey-Bakaikoa, V., Spruce, D., Bowler, M.W. et al. (2010) MxCuBE: a synchrotron beamline control environmentcustomized for macromolecular crystallography experiments. J. Synchrotron. Radiat. 17, 700–707 https://doi.org/10.1107/S0909049510020005

28 Kabsch, W. (2010) Integration, scaling, space-group assignment and post-refinement. Acta Crystallogr. D Biol. Crystallogr. 66(Pt 2), 133–144https://doi.org/10.1107/S0907444909047374

29 Evans, P.R. and Murshudov, G.N. (2013) How good are my data and what is the resolution? Acta Crystallogr. D Biol. Crystallogr. 69(Pt 7), 1204–1214https://doi.org/10.1107/S0907444913000061

30 Collaborative Computational Project Nimber 4. (1994) The CCP4 suite: programs for protein crystallography. Acta Crystallogr. D Biol. Crystallogr. 50(Pt5), 760–763 https://doi.org/10.1107/S0907444994003112

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society 3767

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507

31 Vagin, A. and Teplyakov, A. (2010) Molecular replacement with MOLREP. Acta Crystallogr. D Biol. Crystallogr. 66, 22–25 https://doi.org/10.1107/S0907444909042589

32 Emsley, P. and Cowtan, K. (2004) Coot: model-building tools for molecular graphics. Acta Crystallogr. D Biol. Crystallogr. 60(Pt 12 Pt 1), 2126–2132https://doi.org/10.1107/S0907444904019158

33 Murshudov, G.N., Skubak, P., Lebedev, A.A., Pannu, N.S., Steiner, R.A., Nicholls, R.A. et al. (2011) REFMAC5 for the refinement of macromolecularcrystal structures. Acta Crystallogr. D Biol. Crystallogr. 67(Pt 4), 355–367 https://doi.org/10.1107/S0907444911001314

34 Davis, I.W., Leaver-Fay, A., Chen, V.B., Block, J.N., Kapral, G.J., Wang, X. et al. (2007) Molprobity: all-atom contacts and structure validation forproteins and nucleic acids. Nucleic Acids Res. 35, W375–W383 https://doi.org/10.1093/nar/gkm216

35 Cellini, B., Montioli, R., Paiardini, A., Lorenzetto, A. and Voltattorni, C.B. (2009) Molecular insight into the synergism between the minor allele of humanliver peroxisomal alanine:glyoxylate aminotransferase and the F152I mutation. J. Biol. Chem. 284, 8349–8358 https://doi.org/10.1074/jbc.M808965200

36 Conter, C., Oppici, E., Dindo, M., Rossi, L., Magnani, M. and Cellini, B. (2019) Biochemical properties and oxalate-degrading activity of oxalatedecarboxylase from Bacillus subtilis at neutral pH. IUBMB Life. 71, 917–927 https://doi.org/10.1002/iub.2027

37 Purdue, P.E., Allsop, J., Isaya, G., Rosenberg, L.E. and Danpure, C.J. (1991) Mistargeting of peroxisomal L-alanine:glyoxylate aminotransferase tomitochondria in primary hyperoxaluria patients depends upon activation of a cryptic mitochondrial targeting sequence by a point mutation. Proc. NatlAcad. Sci. U.S.A. 88, 10900–10904 https://doi.org/10.1073/pnas.88.23.10900

38 Beeler, T. and Churchich, J.E. (1976) Reactivity of the phosphopyridoxal groups of cystathionase. J. Biol. Chem. 251, 5267–5271 PMID:845839 Cortijo, M., Jimenez, J.S. and Llor, J. (1978) Criteria to recognize the structure and micropolarity of pyridoxal 50-phosphate-binding sites in proteins.

Biochemical J. 171, 497–500 https://doi.org/10.1042/bj171049740 Deller, M.C. and Rupp, B. (2015) Models of protein-ligand crystal structures: trust, but verify. J. Comput. Aided Mol. Des. 29, 817–836 https://doi.org/

10.1007/s10822-015-9833-841 Ikushiro, H., Hayashi, H. and Kagamiyama, H. (2004) Reactions of serine palmitoyltransferase with serine and molecular mechanisms of the actions of

serine derivatives as inhibitors. Biochemistry 43, 1082–1092 https://doi.org/10.1021/bi035706v42 Brenton, A.G. and Godfrey, A.R. (2010) Accurate mass measurement: terminology and treatment of data. J. Am. Soc. Mass Spectrom. 21, 1821–1835

https://doi.org/10.1016/j.jasms.2010.06.00643 Behnam, J.T., Williams, E.L., Brink, S., Rumsby, G. and Danpure, C.J. (2006) Reconstruction of human hepatocyte glyoxylate metabolic pathways in

stably transformed Chinese-hamster ovary cells. Biochem. J. 394(Pt 2), 409–416 https://doi.org/10.1042/BJ2005139744 Oppici, E., Roncador, A., Montioli, R., Bianconi, S. and Cellini, B. (2013) Gly161 mutations associated with primary hyperoxaluria type I induce the

cytosolic aggregation and the intracellular degradation of the apo-form of alanine:glyoxylate aminotransferase. Biochim. Biophys. Acta 1832, 2277–2288https://doi.org/10.1016/j.bbadis.2013.09.002

45 Asano, N., Ishii, S., Kizu, H., Ikeda, K., Yasuda, K., Kato, A. et al. (2000) In vitro inhibition and intracellular enhancement of lysosomalalpha-galactosidase A activity in Fabry lymphoblasts by 1-deoxygalactonojirimycin and its derivatives. Eur. J. Biochem. 267, 4179–4186 https://doi.org/10.1046/j.1432-1327.2000.01457.x

46 Khanna, R., Soska, R., Lun, Y., Feng, J., Frascella, M., Young, B. et al. (2010) The pharmacological chaperone 1-deoxygalactonojirimycin reduces tissueglobotriaosylceramide levels in a mouse model of Fabry disease. Mol. Ther. 18, 23–33 https://doi.org/10.1038/mt.2009.220

47 Mena-Barragan, T., Garcia-Moreno, M.I., Sevsek, A., Okazaki, T., Nanba, E., Higaki, K. et al. (2018) Probing the inhibitor versus chaperone properties ofsp(2)-iminosugars towards human β-glucocerebrosidase: a picomolar chaperone for gaucher disease. Molecules 23, E927 https://doi.org/10.3390/molecules23040927

48 Trapero, A. and Llebaria, A. (2013) Glucocerebrosidase inhibitors for the treatment of Gaucher disease. Future Med. Chem. 5, 573–590 https://doi.org/10.4155/fmc.13.14

49 Malashkevich, V.N., Strop, P., Keller, J.W., Jansonius, J.N. and Toney, M.D. (1999) Crystal structures of dialkylglycine decarboxylase inhibitorcomplexes. J. Mol. Biol. 294, 193–200 https://doi.org/10.1006/jmbi.1999.3254

50 Schade, S. and Paulus, W. (2016) D-cycloserine in neuropsychiatric diseases: a systematic review. Int. J. Neuropsychopharmacol. 19, pyv102https://doi.org/10.1093/ijnp/pyv102

© 2019 The Author(s). Published by Portland Press Limited on behalf of the Biochemical Society3768

Biochemical Journal (2019) 476 3751–3768https://doi.org/10.1042/BCJ20190507